ADVANCED TECHNIQUES OF INTEGRATION

MITCHELL HARRIS AND JON CLAUS

Foreword. The following is a collection of advanced techniques of integra-

tion for indefinite integrals beyond which are typically found in introductory
calculus courses. While we usually begin working with the general cases, it
might be helpful to construct an example and model it after the general case,
or use one of the exercises at the end of the section as a starting point.

Contents

1. Basic Techniques
We begin with a collection of quick explanations and exercises using standard
techniques to evaluate integrals that will be used later on.

1.1. The Decomposition Method. As we begin using more advanced tech-

niques, it is important to remember fundamental properties of the integral that
allow for easy simplifications. For instance,
Z instead of using some more compli-
cated substitution for something such as (x2 + 1)x dx, we might instead use the
linearity of the integral to write
x4 x2
Z Z Z Z
(x2 + 1)x dx = (x3 + x) dx = x3 dx + x dx = + + C.
4 2
This section has no exercises of its own; instead, be on the lookout for use of this
simple technique in more complex integrals.

1.2. Transforming the Integrand (Simple Substitution). By the chain rule

for derivatives,
Z
u(v(y))0 = v 0 (y) · u0 (v(y)) ⇒ u(v(y)) = u0 (v(y)) · v 0 (y) dy,

where x = v(y). Then the rule allows us to calculate the integral of a less compli-
cated function provided we Zmake an appropriate transformation of the differential.
Consider, for example, x2 (x3 + 1)100 dx. While it is possible to expand the
binomial by the binomial theorem and then use the decomposition method, it is
easier to substitute u = x3 +1. Then the differential should be du = 3x2 dx. Hence,
1
2 MITCHELL HARRIS AND JON CLAUS

Z Z
1 100
x2 (x3 + 1)100 dx = u · (3x2 dx)
3
Z
1
= u100 du
3
1 101
= u +C
303
1
= (x3 + 1)101 + C
303
It is important to remember that u was a variable we made up (to represent
x3 +1) and that it has no meaning to an outside observer. We must always substitute
back in for the original variables, and in this case express our answer in terms of x.

Exercise 1.1. Evaluate the following indefinite integrals by suitably transforming

the integrand.
Z Z
x 1
(1) √ dx (7) 3 dx
Z 1 − x2 (x2+ 1) 2
arctan(x)
Z
1
(2) dx (8) dx
Z 1+ x2 Z x ln x ln(ln(x))
p
x2 · 1 + x3 dx
3
(3) (9) sin(x) sin(x + α) dx
√
Z
x2
Z
(4) x 2 − 5x dx (10) √ dx
Z 2−x
1 Z
1
(5) √ dx (11) dx
Z (1 +
 x) x
x
e 2 + ex
1 1
(6) sin · 2 dx
x x

1.3. Integration by Parts.

1.3.1. Understanding the formula. Given two differentiable functions u and v that
are dependent on x, the product rule for differentiation gives

(uv)0 = u0 v + v 0 u.

Since the indefinite integral is the antiderivative, we can then write

Z Z
0
(uv) dx = (u0 v + v 0 u) dx
Z Z
uv = u0 v dx + v 0 u dx
Z Z
u v dx = uv − v 0 u dx
0

Hence, when the left hand side is hard to evaluate but we can find the integral
on the right hand side, we can apply this formula.
 ADVANCED TECHNIQUES OF INTEGRATION 3

1.3.2. Integrals of Inverses. Applying the integration by parts formula to any dif-
ferentiable function f (x) gives
Z Z
f (x) dx = xf (x) − xf 0 (x) dx.

In particular, if f is a monotonic continuous function, then we can write the integral

of its inverse in terms of the integral of the original function f , which we denote
F (x) + C. Substituting f −1 (x) in the above formula we get
Z Z
f (x) dx = xf (x) − x · (f −1 (x))0 dx
−1 −1

Z
= xf −1 (x) − f ◦ f −1 (x) · (f −1 (x))0 dx

= xf −1 (x) − F ◦ f −1 (x) + C,
which allows for the quick evaluation of the integral of the inverse of a multitude
of functions if we know the integral of the original function.
Exercise 1.2. Evaluate the following indefinite integrals by employing the idea of
integration by parts.
Z Z
(1) xe−x dx (7) sec3 (x) dx
√
Z Z
(2) x ln2 (x) dx (8) (arcsin(x))2 dx
Z Z √
(3) ln(x) dx (9) e x dx
Z Z
(4) arctan(x) dx (10) sin(ln(x)) dx
Z Z
(5) x2 arccos(x) dx (11) eax cos(bx) dx
Z Z p
(6) sec(x) dx (12) ln(x + 1 + x2 ) dx

1.4. Trigonometric Substitutions.

Exercise 1.3. Evaluate the following indefinite integrals using a suitable trigono-
metric substitution.
Z
1 x2
Z
(1) 3 dx (4) √ dx
(1 − x2 ) 2
Z p Z x2 − 2
1
(2) 1 − x2 dx (5) p dx
Z r Z p (x − a)(b − x)
a+x
(3) dx (6) (x − a)(b − x) dx
a−x

1.5. Hyperbolic Substitutions.

Exercise 1.4. Evaluate the following indefinite integrals by using a suitable hy-
perbolic substitution.
4 MITCHELL HARRIS AND JON CLAUS

Z p Z
1
(1) a2 + x2 dx (3) p dx
Z p (x + a)(x + b)
x2
Z
(2) √ dx (4) (x + a)(x + b) dx
a2 + x2

1.6. Review Exercises.

Exercise 1.5. Evaluate the following indefinite integrals using the methods from
this section. (The integrals are not in order of difficulty.)
Z
ln2 (x)
Z
(1) (1 − x)(1 − 2x)(1 − 3x) dx (9) dx
Z x
x+1 Z
√
(2) √ dx (10) x 2 − 5x dx
x
sin x cos2 x
Z Z
1
(3) 2x
dx (11) dx
Z 3x1 + cos Z (x − 1)(x + 3)
e +1 1
(4) x+1
dx (12) 3 dx
e (x + a2 ) 2
2
√
Z Z
3
(5) 1 − 3x dx (13) xn ln x dx (n 6= −1)
Z x+1
2 − 5x−1
Z p
(6) x
dx (14) x2 a2 + x2 dx
Z r 10 √
x−a x ln(x + 1 + x2 )
Z
(7) dx (15) √ dx
Z x+a 1 + x2
1
Z
(8) dx (16) (ex − cos x)2 dx
1 + cos x

2. “Rational” Functions
2.1. Denominator is Quadratic. The simplest rational function with a quadratic
1
denominator is . If we try to evaluate this integral, it is helpful to
ax2 + bx + c  
b 2 b2


“complete the square” and write ax2 + bx + c = a x + 2a + c − 4a . Then it is
b2
easier to see that for c − 4 >0
Z Z
dx dx
=
ax2 + bx + c b

2 q
b2

2
a x + 2a + c− 4a
Z
1 dx
=
a u2 + v 2
1 u
= arctan + C,
av v
q b
b c− 2a
where u = x + 2a and v = a .
b2
In the case that c − ≤ 0, then the denominator can be factored and evaluated
4
using partial fraction decomposition.
This technique can be applied to integrals where the situation may not be so
obvious. For instance, for the integral
Z
(x + 1)
2
dx
x +x+1
 ADVANCED TECHNIQUES OF INTEGRATION 5

one might try an immediate substitution, which would fail. First split the numer-

2
ator x + 1 = x + 12 + 12 to match the denominator written as x + 12 + 12 . Now

it is more obvious how to apply the above technique along with a substitution as

x + 21 1
Z Z Z
(x + 1) 2
2
dx = 2 dx +
2 dx
x +x+1 x + 21 + 12 x + 12 + 12


 
1 1 2x + 1
= ln(4x2 + 4x + 3) + √ tan √ +C
2 2 2

2.2. Quadratics Beneath Radicals. For integrals of the form

Z
dx
√
2
ax + bx + c
we combine the ideas of completing the square with the trigonometric substitution
x = a tan θ (if a > 0) or x = a sin θ (if a < 0).

2.3. Ostrogradsky’s Method.

Exercise 2.1. Evaluate the following indefinite integrals by using Ostrogradsky’s

Method.
Z Z
x 1
(1) 2 3
dx (4) dx
Z (x − 1) (x + 1) (x4
+ 1)2
1 x2
Z
(2) 3 2
dx (5) dx
Z (x + 1)
2 2
1 Z (x + 22x + 2)
(3) dx x + 3x − 2
(x + 1)3
2 (6) dx
(x − 1)(x2 + x + 1)2

2.4. Powers of Polynomials.

Z
1
Exercise 2.2. Find a recurrence relation to evaluate the integral In = 2 + bx + c)n
dx,
Z (ax
1
where a 6= 0. Use that result to evaluate I3 = dx. The identity
(x2 + x + 1)3
2 2 2
4a(ax + bx + c) = (2ax + b) + (4ac − b ) should be useful.
x+a
Exercise 2.3. Use the substitution t = x+b to evaluate the given indefinite inte-
grals, where m and n are integers.
Z Z
1 1
(1) dx (2) dx
(x − 2)2 (x + 3)2 (x + a)m (x + b)n

2.5. Taylor’s Formula.

Exercise 2.4. If Pn is some polynomial

Z of degree n, then by use of Taylor’s formula,
Pn (x)
evaluate the indefinite integral dx.
(x − a)n+1
6 MITCHELL HARRIS AND JON CLAUS

2.6. Transformation to Rational Function. When the denominator of a frac-

tion is not rational, we cannot apply the methods above directly to evaluate its
indefinite integral. First, we make a change of variables that eliminates the irra-
tional factors.
For instance, to evaluate Z
1
√ dx,
1+ x
there√is no partial fraction decomposition that will help. Therefore we first take
u = x so that
1 √
du = √ dx ⇒ dx = 2 x du = 2u du.
2 x
Z
2u
The original indefinite integral is hence equal to du. We can proceed as in
1+u
the previous methods by dividing the numerator into the denominator.
Exercise 2.5. Evaluate the following indefinite integrals by transforming the in-
tegrand to a rational function by a suitable change of variables.
Z Z
1 1
(1) √ √ dx (4) p dx
x(1√+ 2 x + x) 3 3
(x + 1)2 (x − 1)4
x32+x
Z Z
1
(2) √3
dx (5) √ √ dx
x + √ 2 + x 1 + x + 1+x
1− x+1 1 − x + x2
Z Z
(3) √ dx (6) √ dx
1+ 3x+1 1 + x − x2

2.7. Rational Functions

√ with Irrational Factors. If Pn (x) is a polynomial of
2
degree n, and if y = ax + bx + c, then there exist some real number λ and n − 1
degree polynomial Qn−1 (x)such that
Z Z
Pn (x) dx
dx = Qn−1 (x)y + λ .
y y
We may verify this result by writing in the definition for y and differentiating to
get that
Pn (x) p 1
√ = Qn−2 ax2 + bx + c + λ √ ⇒ Pn (x) = Qn (x) + λ
2
ax + bx + c 2
ax + bx + c
for some other nth degree polynomial, Qn (x).
Z
P (x)
Exercise 2.6. Generalize the above formula for the indefinite integral, dx.
Q(x)y
P (x)
(Hint: Consider the partial fraction decomposition of Q(x) .)

Exercise 2.7. Using the above ideas or another method, evaluate the following
indefinite integrals.

x2 a1 x2 + b1 x + c1
Z Z
(1) √ dx (3) 2
dx
1 + 2x − x2 Z ax + bx + c
Z
x2 x
(2) √ dx (4) √ dx
1 + x + x2 (x − 1) 1 + 2x − x2
2
 ADVANCED TECHNIQUES OF INTEGRATION 7

Z
1 Z √
(5) √ dx x2 + x + 1
(7) dx
(1 − x)21− x2 (x + 1)2
Z √ 2
x + 2x + 2x
(6) dx

2.8. Fractional Linear Transformations + Euler’s Substitutions. Euler found

that for integrals involving square roots of quadratics, the following substitutions
are sometimes helpful.
√ √
• If a > 0, then set √ax2 + bx + c = ± ax + z for some suitable choice of z.
√
• If c > 0, then set p ax2 + bx + c = xz ± c for some suitable choice of z.
• Or sometimes take a(x − x1 )(x − x2 ) = z(x − x1 ).
Exercise 2.8. Evaluate the following indefinite integrals by using an Euler Sub-
stitution.
Z Z
1 p
(1) √ dx (3) x x2 − 2x + 2 dx
Z x+ x2 + x + 1
1
(2) √ dx
1 + 1 − 2x − x2

2.9. Chebyshev’s Theorem. The integrals we have treated thus far have either
been square roots of functions or functions raised to integral powers. The indefinite
integral of many functions with other rational exponents cannot be written in terms
of elementary functions. In particular,
Z
xm (a + bxn )p dx

with m, n, p rationals can be reduced to the integral of a rational function only in

the following three cases:
• p is an integer: Set x = z N , where N is the least common denominator
of the fractions m and n.
• m+1
n is an integer: Set a + bxn = z N , where N is the denominator of the
fraction p.
−n
• m+1
n + p is an integer: Set ax + b = z N , where N is the denominator
of the fraction p.
This result is due to Chebyshev. We will motivate the substitutions in each of
the given cases; however, the proof that these are the only cases in which there
exists a clever substitution is beyond the scope of this manual.

2.9.1. Case p is an integer. When p is an integer, we could continue with previous

techniques if the integral were transformed so that m and n were integers. Taking
0
z N = x with N the least common denominator of m and n gives xm = (z N )m = z m
0
for some integer m0 . Similarly, xn = (z N )n = z n . Also, dx = N z N −1 d dz and so
the integral
Z
0 0
N · z m +(N −1) (a + bz n )p dz

can be evaluated by a multitude of methods, including expanding by the binomial

theorem and using the decomposition method.
8 MITCHELL HARRIS AND JON CLAUS

2.9.2. Case m+1n is an integer. Since p is not necessarily an integer, we focus the
substitution on the binomial term. If the binomial were raised to the power of
the denominator of p, say N , then N p is an integer. So let z N = a + bxn , where
 N m n
xm = z b−a . This substitution works because N z n−1 dz = bnxn−1 dx, so the
integral becomes

m
(z N − a)/b n
Z
N −1
n−1 · z · z N p dz.
N
((z − a)/b) n

Furthermore, m n − n−1
n is an integer if m−n+1
n = m+1n − 1 is an integer. Since this
is that case in particular, this substitution works to make all the powers integers
and we may continue as before.
−n
2.9.3. Case m+1
n +p is an integer. The verification that the substitution ax +b =
N
z works is similar to the previous case but a bit more technical. It is left as an
exercise to the interested reader.
Exercise 2.9. Evaluate the follow indefinite integrals using Chebyshev’s theorem.
Z p Z
1
(1) 3 4
x + x dx (3) √4
dx
4
Z Z p1 + x
1 3
(2) √
3
dx (4) 3x − x3 dx
1 + x3

3. Trig Integrals
We will assume competency with the integrals of common trigonometric expres-
sions.
3.1. “Rational” Trig Functions. When confronted with an integral of the form
Z
A sin x + B cos x
dx,
a sin x + b cos x
d
we need to exploit the relationship that dx (a sin x + b cos x) = a cos x − b sin x. We
can write the numerator as a linear combination c1 (a sin x + b cos x) + c2 (a cos x −
b sin x) in order to use the decomposition method to evaluate the integral as
u0
Z  
c1 + c2 dx, with u = a sin x + b cos x,
u
which evaluates to c1 x + c2 ln |a sin x + b cos x| + C. The constants c1 and c2 are
determined by the equation A sin x + B cos x = c1 (a sin x + b cos x) + c2 (a cos x −
b sin x). Comparing coefficients of the sine and cosine terms gives us a system of
equations that can be solved to get
aA + bB aB − bA
c1 = 2 and c2 = 2 .
a + b2 a + b2
In contrast, integrals of the form
Z
1
dx
a sin x + b cos x
do not have the numerator to allow for helpful cancellation. We know that a linear
combination of trig functions is in turn a simple trig function. Our goal is to turn
the denominator into an expression of the form A sin(x+φ) for some constant phase
 ADVANCED TECHNIQUES OF INTEGRATION 9

angle φ. Using the angle sum formula, this is A cos φ sin x + A sin φ cos x. We can
set this equal to the original denominator,
a sin x + b cos x = (A cos φ) sin x + (A sin φ) cos x
√
and equate the coefficients as a = A cos φ and b = A sin φ. Then A = a2 + b2 and
tan φ = ab . So all we need to do is calculate
−1
Z Z
1 1
dx = csc(x + φ)dx = ln | csc(x + φ) + cot(x + φ)| + C
A sin(x + φ) A A
and substitute in for A and φ.
3.2. Hyperbolic Functions. When dealing with integrals of hyperbolic functions,
there are two general ways to proceed. We can either evaluate them with the
substitution ex = u or using hyperbolic identities, such as cosh2 x − sinh2 x = 1.
The substitution method is more general and builds on previous sections. This
is because all hyperbolic functions are rational functions of ex , meaning that there
exists some rational function f (ex ) = h(x), where h(x) is a hyperbolic function.
In addition, the substitution maps dx to u1 du. Hence, after the susbtition, any
hyperbolic function will be transformed into an expression involving polynomials
of u in place of the hyperbolic functions.
An alternative to the substitution method is to evaluate hyperbolic integrals in
the same manner as trignometric integrals, but with a slightly different identity set.
Here is one such example:

sinh(2x) − 1
Z Z
sinh2 (x)dx = dx)
2
sinh(2x) x
= − +C
4 2
This following example is evaluated using the standard substitution.

u4
Z Z  
2 2 1 1 1
sinh (x) cosh (x)dx = − − dx)
16 16u4
u 8
u4 1 ln u
= + 4
− +C
64 64u 8
e4x e−4x x
= + − +C
64 64 8
sinh(4x) x
= − +C
32 8
In general, it is often hard to convert a polynomial of ex back into hyperbolic
form, which is one of the shortcomings of the substitution method. One of the
strengths of this method is that it allows any rational function of hyperbolic func-
tions to be evaluated using the substitution and Ostrogradsky’s Method.
Lastly, one can convert a hyperbolic function using the relations i sinh(x) =
sin(ix) and cosh(x) = cos(ix) and the substitution ix = u. However, this involves
integration of complex numbers and will not be dealt with in this text.
Exercise 3.1. Evaluate the follow indefinite integrals containing hyperbolic func-
tions.
10 MITCHELL HARRIS AND JON CLAUS

Z Z p
(1) cosh4 (x) dx (5) tanh(x) dx
Z Z
1
(2) sinh3 (x) dx (6) dx
Z Z sinh(x) + 2 cosh(x)
cosh(x)
(3) sinh(x) · sinh(2x) · sinh(3x) dx (7) dx
Z
1 Z 3 sinh(x) − 4 cosh(x)
(4) dx
(8) sinh(ax) · cosh(bx) dx
sinh2 x − 4 sinh x cosh x + 9 cosh2 x

Acknowledgements
The exercises themselves were largely translated and selected from the Russian
text of Gradshteyn Ryzhik. The authors would like to thank Sergei Shabanov for
his help with the translation and other helpful comments. Please report errors to
[email protected].